Applied Thermal Engineering 67 (2014) 201e213

Contents lists available at ScienceDirect

Applied Thermal Engineering

journal homepage: www.elsevier.com/locate/apthermeng

Numerical modeling of flow through an industrial burner orifice

L.C.B.S. Reis a, b, c, J.A. Carvalho Jr. a, *, M.A.R. Nascimento d, L.O. Rodrigues d, F.L.G. Dias d,
P.M. Sobrinho a
a
Universidade Estadual Paulista, Campus de Guaratinguetá, Guaratinguetá, SP, Brazil
b
Fundação Oswaldo Aranha, Volta Redonda, RJ, Brazil
c
Usina Presidente Vargas, Companhia Siderúrgica Nacional, Volta Redonda, RJ, Brazil
d
Universidade Federal de Itajubá, Itajubá, Brazil

h i g h l i g h t s

Numerical modeling of natural gas flow through an industrial burner was performed.

Standard, RNG, Realizable keε, and Reynolds Stress Model (RSM) have been used.

The considered models represent the experimental conditions.

a r t i c l e i n f o a b s t r a c t

Article history: This paper presents numerical modeling of a turbulent natural gas flow through a non-premixed in-
Received 7 November 2013 dustrial burner of a slab reheating furnace. The furnace is equipped with diffusion side swirl burners
Accepted 14 February 2014 capable of utilizing natural gas or coke oven gas alternatively through the same nozzles. The study is
Available online 4 March 2014
focused on one of the burners of the preheating zone. Computational Fluid Dynamics simulation has
been used to predict the burner orifice turbulent flow. Flow rate and pressure at burner upstream were
Keywords:
validated by experimental measurements. The outcomes of the numerical modeling are analyzed for the
Industrial combustion
different turbulence models in terms of pressure drop, velocity profiles, and orifice discharge coefficient.
Process simulation
Numerical analysis
The standard, RNG, and Realizable keε models and Reynolds Stress Model (RSM) have been used. The
Industrial burner main purpose of the numerical investigation is to determine the turbulence model that more consis-
Natural gas tently reproduces the experimental results of the flow through an industrial non-premixed burner
orifice. The comparisons between simulations indicate that all the models tested satisfactorily and
represent the experimental conditions. However, the Realizable keε model seems to be the most
appropriate turbulence model, since it provides results that are quite similar to the RSM and RNG keε
models, requiring only slightly more computational power than the standard keε model.
Ó 2014 Elsevier Ltd. All rights reserved.

1. Introduction burnt through continuous diffusion only after being discharged

The search for higher energy efficiency of industrial scale com-
bustion furnaces and burners to obtain fuel savings has demanded
experimental studies that are complemented with Computational
Fluid Dynamics simulations. Industrial burners operate in non-
premixed conditions for safety reasons. Fuel and oxidizer enter
separately into the combustion furnace and are then mixed and
* Corresponding author. Av. Ariberto Pereira da Cunha, 333, Departamento de
Energia, Guaratinguetá, SP CEP 12516-410, Brazil.
E-mail addresses: joao@feg.unesp.br, joao.a.carvalho.jr@pesquisador.cnpq.br
(J.A. Carvalho).
from the orifice. Gas fuel is supplied through the orifice and com-
bustion air enters from the surroundings so that the gas can only be
burnt within a certain distance from the orifice. Burner designers
must determine the correct area of the fuel orifice. For a deter-
mined orifice size, there is a specific fuel pressure in order to have
the right mix with the combustion air. High fuel pressure can result
in soot or flame impingement and low fuel pressure can result in air
excess and may not allow the achievement of the required furnace
heat load.
Combustion modeling studies have been intensely developed
during the last 20 years and the phenomena of flame diffusion is
usually treated downstream from the orifice burner. Great

http://dx.doi.org/10.1016/j.applthermaleng.2014.02.036
1359-4311/Ó 2014 Elsevier Ltd. All rights reserved.
202 L.C.B.S. Reis et al. / Applied Thermal Engineering 67 (2014) 201e213
Nomenclature
A geometrical orifice area (m2)
c speed of sound (m s1)
Cd discharge coefficient of the orifice (dimensionless)
d nozzle diameter (m)
D pipe diameter upstream of the nozzle (m)
f friction factor (dimensionless)
K pressure drop coefficient for losses of fittings
(dimensionless)
k turbulent kinetic energy per unit mass (m2 s2)
k ratio of specific heats of the fuel (dimensionless)
L length of pipe (m)
m_ mass flow rate (kg s1)
MW molecular weight (kg kgmol1)
P1 fuel pressure at upstream from the nozzle (N m2)
P2, Pb fuel pressure at downstream from the nozzle (N m2)
Q volumetric flow rate (m3 s1)
R universal gas constant (J kgmol1 K1)
T temperature (K)
attention has been given to investigating the interaction of turbu-
lence combustion and its consequences, including deciding which
turbulence model is the most appropriate depending on the spec-
ificity of the application. Usually the burner orifice modeling does
not take part of the analysis. However, for industrial burners it is
very important to have information about the pressure upstream
from the burner, since changes in fuel chemical composition can be
compensated by altering fuel pressure. Burner nozzles are designed
for a certain orifice discharge coefficient obtained from experi-
mental measurements. This coefficient is defined as the ratio of
actual flow to the maximum theoretical flow and is normally ob-
tained using empirical correlations based on experimental data,
derived under controlled laboratories conditions.
The present paper reports a numerical analysis identifying the
most appropriate turbulence model to simulate the flow through
an industrial burner orifice, establishing comparisons with the best
turbulence models for some combustion applications. Flow rate
and pressure at burner upstream were validated by experimental
measurements. The results indicate that the standard, RNG, and
Realizable keε models and Reynolds Stress Model (RSM) satisfac-
torily represent the experimental conditions. However, the Real-
izable keε model seems to be the most appropriate turbulence
model, since it provides results that are quite similar to the RSM
and RNG keε models, requiring only slightly more computational
effort than the standard keε model. The industrial burner is a non-
premixed swirl burner from the preheating zone of slab reheating
furnaces. It is from Usina Presidente Vargas, a major steelworks that
belongs to Companhia Siderúrgica Nacional (CSN), located in the
state of Rio de Janeiro, Brazil.
2. Literature review
Few attempts have been made to simulate turbulent combus-
tion with burner orifice flow pattern. Indeed, most of the numerical
and experimental works have studied orifice meters, and a few
have studied hydraulic orifices. The following presents a list of
published literature of correlated papers. The quality of the simu-
lations for orifice flow and the choice of turbulent models are
discussed.
Experiments on orifice fluid flow measurements were simulated
[1] using data provided by others [2,3]. CFD simulations were
validated through pressure drop and energy balance
Greek symbols
a port angle in a burner nozzle
b ratio of orifice diameter to pipe diameter
D difference in a quantity
ε turbulent energy dissipation
Subscripts
1 nozzle entrance
2 nozzle exit
Abbreviations
ICEM CFD a meshing software
CFD computational fluid dynamics
DO discrete ordinate method
EDM Eddy-Dissipation Model
LFM Laminar Flamelet Model
PDF Probability Density Function
RSM Reynolds Stress Model
WI Wobbe Index
WSGGM weighted sum of gray gases model
measurements using water as fluid. The standard keε turbulence
model was used. A comparison of the numerical and experimental
results revealed that experimental data closely agreed with CFD
predictions.
Experimental discharge coefficients for flow meters, including
orifice plate flow meters, were obtained in order to validate
numerical results at low Reynolds numbers [4]. The Realizable keε
model was used for turbulence closure. The intent of the study
was to present characteristic curves to enable users to better
understand the relative differences expected at low Reynolds
numbers.
CFD was applied to numerically predict the calibration coeffi-
cient of orifice meters in order to ease the laborious experimental
procedure of calibration [5]. The methodology satisfactorily pre-
dicted the discharge coefficients.
The flow through a circular orifice was investigated using CFD
and two turbulence modeling techniques [6]. The standard keε
model and the Reynolds Stress Model (RSM) were employed. It was
found that the results agreed well with experimental data. How-
ever, the RSM was more accurate in the downstream orifice region
than the keε model.
A numerical study was conducted to evaluate effects of flow
through a simple orifice [7]. The main recommendations are that
the grid spacing must be 0.1% of pipe diameter upstream of the
plate and the use of high-order differencing schemes in order to
calculate pressure loss correctly. The standard keε model was used.
It agreed sufficiently with experimental data, but the authors rec-
ommended the use of other turbulence models or modification of
the keε model to improve performance.
Another study [8] proposed to relax the square root relation
commonly used by international standards to determine the flow
rate through a specific discharge coefficient value. The resulting
power law relation was shown to improve accuracy. In addition to
the experimental data, evidence was also obtained by performing
numerical simulations. The standard keε turbulence model was
used.
A single empirical formula to model the flow through hydraulic
orifices was proposed [9]. It makes use of a linear relation for small
pressure differences and the conventional square root law for tur-
bulent conditions. Simulation results have proved to be accurate.
The effect of contaminated orifice plates on the discharge
coefficient was investigated [10]. Experimental work was
 L.C.B.S. Reis et al. / Applied Thermal Engineering 67 (2014) 201e213
conducted simulating contamination by sticking circular metal
discs. Computational fluid dynamics was used to assist interpre-
tation and had good agreement with experiments.
Several works have been made employing computational fluid
dynamics simulations of flow in non-premixed burners. Those
works were made to investigate or predict flame properties and
combustion flow fields [11], pollutant emissions [12e19] and the
mechanism of flame stability [13]. Combustion modeling has been
performed using several turbulence models, the standard keε
[11,14e22], Reynolds Stress Model [11,12,16,22] and Realizable keε
[14,21]. With respect to combustion models for non-premixed
combustion for conventional flames, the most used are the Prob-
ability Density Function/Laminar Flamelet Model (PDF/LFM)
[12,23] and the Eddy Dissipation Model (EDM) [1,13,15,17,18,20e
22]. The Eddy Dissipation Concept (EDC) [14,16] is mainly used
for flameless combustion. EDM was chosen for this study. The
quality of combustion modeling simulations using EDM and the
choice of turbulence models are discussed in the following.
Modeling of non-premixed swirl burner flows was investigated
using RSM and standard keε turbulence models [11]. The results
agreed well with experimental data. Results of computational and
experimental investigations of a turbulent asymmetric vortex
flame were presented [13]. The three-dimensional reacting flow
fields were described using EDM and the modified standard keε
turbulence model, called Rε/keε. The mechanism of flame stability
and interaction with the forced vortex field was discussed. The
effect of increasing the intensity of turbulence in the air stream on
NOx and soot formation in turbulent methane diffusion flames was
investigated [18]. The interaction between turbulence and com-
bustion in the flame field was modeled using the standard keε and
EDM models. It was found that increased intensity of free stream
turbulence in the air supply results in a significant reduction in NO
formation. Reduction of soot formation was also found. The
implementation of EDM in an OpenFOAM CFD toolbox was dis-
cussed [20]. The code was validated by modeling a confined non-
premixed methane jet flame using the standard keε turbulence
model. The predictions were compared with published experi-
mental results and the ANSYS Fluent predictions. The results
showed that the predictions agreed with the experimental results.
The performance of four turbulence models in modeling a co-
flowing turbulent diffusion methane/air flame based on EDM was
analyzed [21]. The method of computing the dissipation rate of
turbulent kinetic energy is considered crucial to accurately express
reaction rate. It was shown that the Realizable keε turbulence
model is better than the other models, including the standard keε
model. A numerical study of the swirl effect on a coaxial jet com-
bustion diffusion flame was presented [22]. EDM and the standard
keε turbulence model were used. The results showed that EDM
agreed well with experimental data.
3. Experimental setup
The primary function of the hot rolling furnace is to reheat semi-
finished steel slabs to temperatures of about 1250  C. The interior of
the furnace is divided into seven zones for temperature control:
preheat, top-and-bottom; heating, top-and-bottom; and soak, top-
and-bottom, with two zones for top soaking. All zones can burn
natural gas or coke oven gas with preheated combustion air by
means of recuperative heat exchangers. The preheating zone of
the reheating furnace is equipped with six side swirl burners,
three on each side of the furnace. Each burner has a thermal
power of 4360 kW. The walking-beam reheating furnace has inner
dimensions of 34,806  11,580  3,800 mm3 (length  width 
height).
203
Fig. 1 shows a longitudinal cross-section view of the furnace
with an indication of the burner used in this study. The skids are not
shown in this view. The computational domain considered is
crosshatched in Fig. 1. Its perspective view is detailed in Fig. 2a. It
comprises the region surrounding the flame up to the limits of the
neighbor burner at right, which defines a surface of flue gas outlet
in the direction of the natural draft stack; the incoming flue gases
from bottom heating and bottom soaking zones are at the left; the
refractory floor is at the bottom; the slab is at the top; the centerline
of the furnace is seen as a symmetrical surface, since there is an
opposite burner in the same direction; and the casing/refractory
wall which is equipped with the burner is shown.
Inside the furnace, the slabs are supported by water-cooled,
refractory-coated pipes called “skids”. To minimize the cold spots
or “skid marks” left in the slab, the skid spacing changes twice
along the furnace. This is the reason for the variation of direction of
the skid pipe. There are two independent set of skids, one fixed and
one moving, which take turns supporting the slab as it is trans-
ported through the furnace by a frame moved by large hydraulic
cylinders. The skids also are represented in the numerical model as
shown in Fig. 2a. Thus, there are two boundaries of heat absorption:
one the skid cooling tubes; and other the slab on the top of the
domain. Fig. 2b shows a photograph of the burner portion outside
of the furnace.
The BHF.M.10-type swirl configuration preheating zone burners
were manufactured by Stein Heurtey. Each burner has eight gas
nozzles with diameters of 33 mm, twelve primary air nozzles with
diameters of 59 mm which promote air swirl, sixteen secondary air
nozzles with diameters of 51 mm, and a heavy oil lance which is not
in operation anymore. The burner geometry created on ICEM CFD is
shown in Figs. 3a, b and 4a. Fig. 4b shows a side view of the burner.
The furnace operates at atmospheric pressure. Furnace pressure
was measured at the boundary surface of the outlet gases using an
LD 3000 M pressure gauge (accuracy  0.04%; manufacturer: Beta
Calibrator-Martel Electronics). Furnace temperatures were
measured with type R thermocouples (accuracy  1.8  C; manu-
facturer: Ecil). Air and gas pressure were measured by betagauge
320 pressure calibrator (accuracy  0.075%; Martel Electronics) and
Testo 521-1 (accuracy  0.2%; manufacturer: Testo). Air flow rate
was measured by LD 301-D1 (0.04%; manufacturer: Smar) and gas
flow rate was measured by LD 301-D3 (0.04%; manufacturer:
Smar). Temperature inside the furnace was measured by a ther-
mocouple on the end of a stainless steel lance of 6 m length, placed
through the heavy oil nozzle (centerline of the burner) and posi-
tioned at 4.5 m distance from the burner wall.
The experimental data that have been taken for the present
simulations are considered to be closest to steady state operation.
Natural gas was used as the fuel. Its main characteristics are pre-
sented on Table 1.
4. Theory
The review of the literature demonstrated that there is little
information about turbulence numerical analysis to simulate the
flow through an industrial burner orifice. Consequently, this study
was performed to analyze the available turbulence models for this
type of application.
In order to simulate the steady state flow though the nozzles,
the Reynolds-averaged NaviereStokes equations were solved,
including mass conservation, momentum conservation, turbulent
kinetic energy and energy dissipation rate equations. Although the
domain of analysis is upstream from the beginning of chemical
reaction, the complete domain also took the zones with reaction
flows into account in order to assure that mass and energy balances
were consistent. Conservation for the species transport equation
204 L.C.B.S. Reis et al. / Applied Thermal Engineering 67 (2014) 201e213
Fig. 1. Longitudinal cross-section of the reheating furnace showing the computational domain.
was considered as well. Those equations were solved for steady
state conditions by utilizing FLUENT computational fluid dynamics
[24].
5. Computational work
Three dimensional simulations have been carried out for
modeling. Several operational conditions were considered, varying
the flow rate of natural gas and, consequently, of the combustion
air. A furnace sketch showing the polyhedral mesh applied for the
whole domain is presented in Fig. 5a and a detailed mesh view of
the burner nozzles is shown in Fig. 5b.
Pressure-velocity coupling was accomplished by using SIMPLE
(Semi Implicit Pressure Linked Equation) algorithm. The solver
used was the “pressure-based segregated”, in which the governing
equations are solved in sequence. Interpolation of variables to cell
faces was performed using a second-order up-winding scheme,
except for pressure, which was used the “standard” scheme. The
under-relaxation parameters were set to 0.3 for pressure, 0.6 for
density, 0.7 for body forces, 0.7 for momentum equations, 0.5 for
turbulent quantities, 0.2 for energy and 0.3 for chemical species.
A computational mesh was created for the domain, as shown in
Fig. 5a. A refined mesh was used for the burner nozzles as shown in
Fig. 5b. Mesh geometries contained on the order of 650,000 cells. A
three dimensional tetra-mixed mesh was generated using ICEM
CFD and converted to polyhedral cells in FLUENT 12.0.16, the
commercial code used to perform the simulations. The skid cooling
tubes were entirely incorporated into the simulations. The calcu-
lations were performed on a computer with 08 Intel Xeon Quad-
core 5420 processors (2.5 GHz/12 MB) and 16 GB RAM, connected
to a cluster with 40 cores distributed per a total of 5 computers.
5.1. Turbulent combustion modeling
The standard keε turbulence [25] was used. It is widely vali-
dated and it has also been shown to have excellent performance for
many relevant industrial flows. However, as the case in point is a
swirl burner and the most recommended turbulence modelings in
swirling flows are the RNG keε, the Realizable keε and the Rey-
nolds stress models, they were also used in the simulation.
Combustion modeling used the Eddy-Dissipation (EDM) model,
which considers fast burning, with the overall rate of reaction
controlled by turbulent mixing. For cases of non-premixed flames,
turbulence slowly convects and mixes fuels and oxidizer into the
reaction zones, where they burn very quickly. Combustion proceeds
whenever turbulence is present and every reaction has the same
turbulent rate, which is used only for one-step or two-step global
reactions. Thus, the limitation of the model is that kinetically
controlled species as radicals cannot be predicted and intermediate
species are neglected. The present work does not concern the
prediction of such radicals and the problem being modeled is
clearly a fast burning case of high temperature flame, so the EDM is
an appropriate model to be used, avoiding the use of multi-step
chemical mechanisms based on Arrhenius rates, which can differ
for each reaction. Radiative heat transfer was accounted for using
the discrete ordinates (DO) method. The absorption coefficient of
the flue gas used the domain-based weighted sum of gray gases
(WSGGM) model.
Properties of the gas mixture were calculated as follows: density
by the ideal gas law and compared to constant density; heat ca-
pacities through the mixing law, determined per species through a
piecewise-polynomial, enabling composition and temperature
dependence. This condition is considered very important for
combustion applications to correctly predict flame temperature,
since it substantially reduces the peak temperature; molecular
viscosity and thermal conductivity by ideal gas mixing law; and
mass diffusivity determined by constant dilute approximation co-
efficient of 2.88  105 m2/s.
5.2. Mesh analysis
A mesh study was performed in order to be sure it was suffi-
ciently refined. The results of important parameters such as tem-
perature along the furnace and pressure and velocity through the
burner nozzle were analyzed for two different grid sizes. The coarse
grid size has 501,853 cells. The fine grid size has 650,000 cells, all
the additional cells in the burner domain. The values of the
analyzed variables did not change significantly, as shown in Fig. 6,
despite a slight reduction in the upstream nozzle pressure for the
fine grid (Fig. 6b). This result was very similar to the centerline
pressure profile for an orifice meter [1]. The difference with respect
to the curve shape is due to the greater length of the orifice for the
burner nozzle, 125 mm, which causes a less abrupt pressure drop
along its orifice.
 L.C.B.S. Reis et al. / Applied Thermal Engineering 67 (2014) 201e213 205

Fig. 2. A perspective view of the computational domain and its boundaries (a), and a front view of the burner outside the furnace (b).

Wall function approach was used to model the near-wall region,
since it gives reasonable predictions for high-Reynolds-number
wall-bounded flows, which is the case in this work. “Standard
wall functions” and “Non-equilibrium wall functions” were used
with the aim of comparing the results between those methods,
considering that the latter is more appropriate for flows that are
characterized by significant pressure gradients, since strong gra-
dients occur in the region of the nozzles. However, the results were
the same for both methods.
The dimensionless distances from the wall were checked for
each type of nozzle, considering several burner thermal
loads. Table 2a presents the results of yþ for the coarse grid and
Table 2b for the fine grid.
Wall roughness effects were taken into account in the turbu-
lent wall-bounded flow. The law of the wall was modified to
consider the rough surface of the burner nozzle, as available on
FLUENT. It is based on experiments with roughened pipes which
indicate that the mean velocity distribution near rough walls has
the same slope, but a different intercept, i.e., an additive constant
in the log-law. Therefore, a term was introduced in the law-of-
the-wall, which is correlated with the physical roughness
height. In order to preserve physical significance, the distance
206 L.C.B.S. Reis et al. / Applied Thermal Engineering 67 (2014) 201e213
Fig. 3. Perspective view of the BHF.M.10-type burner: external side (a), furnace side
(b).
from the wall to the centroid of the wall-adjacent cell was greater
than the roughness height.
5.3. Boundary conditions
The domain is made up of the regions shown in Figs. 1 and 2a.
The burner orifices are inlet flow boundaries, including fuel, pri-
mary and secondary combustion air nozzles. The boundaries inside
the furnace are: the refractory floor classified as a wall; the top is
the slab, another wall; the symmetry wall, which represents the
furnace centerline; flue gas inlet flow boundary is the incoming
gases from the bottom heating and soaking zones of the furnace;
and flue gas outlet flow which is the sum of the gases from the
burner itself and the flue gas inlet flow from the bottom heating
and soaking zones. Table 3 summarizes the boundary conditions
and the results of mass and energy balances for six different burner
heat loads, related to CFD simulations.
5.3.1. Heat transfer at wall boundaries
The slab is a wall boundary at the top of the domain. The cal-
culations were made using a fixed heat flux boundary condition. An
average heat flux transmitted to the slab was considered based on
the reheating furnace software, which provides the bottom surface
and core slab temperature for each axial position inside the furnace,
allowing the heat flux to be determined by conduction. The
predicted bottom slab temperature was compared to the value
provided by the software. The mentioned software is a supervisory
system for furnace combustion optimization, provided by the
manufacturer Stein Heurtey for furnace operation.
The skid cooling tube was also considered a fixed heat flux
boundary condition. Average heat flux transmitted to the skid was
based on the total heat absorbed by this system for the entire
furnace, which is calculated by measuring the cooling water flow
rate and inlet and outlet temperatures.
6. Results and analysis
6.1. Pressure profiles for natural gas
The results are presented in the form of fluid pressure upstream
the burner nozzle vs. fluid flow rate. Primarily for natural gas,
pressure upstream from the burner nozzle was measured and
compared with the numerical predictions, as shown in Fig. 7. The
predicted pressure profiles for all turbulence models agreed well
with the experimental measurements. However, the results with
the standard keε turbulence model indicate that the predicted
pressure is somewhat larger than the other turbulence models.
Similarly, primary combustion air pressure experimental mea-
surements agreed well with numerical calculations, as shown in
Fig. 8. Likewise, the results for standard keε model indicate values
slightly higher than the other models. Last, the comparison be-
tween secondary combustion air pressure measurements and nu-
merical results also agreed well; however, all turbulence models
resulted in identical figures, including the keε model as shown in
Fig. 9. This may be attributed to the influence of the swirling flow in
the natural gas and primary combustion air nozzles. The secondary
combustion air nozzle has less influence on swirling flow, as can be
seen in Fig. 10a, b and c, which show the calculated streamlines in
the downstream burner region. Fig. 10a shows the calculated
streamlines released from natural gas nozzles; Fig. 10b shows the
streamlines from primary combustion air nozzles; and Fig. 10c
shows the streamlines from secondary combustion air nozzles. As
RSM, RNG keε and Realizable keε models account for the effects of
swirl, they have greater potential to predict more accurate pre-
dictions for swirling flows.
Considering computation time, memory needed and number of
iterations to converge, the Realizable keε model is the most
appropriate turbulence model to be used for the present applica-
tion, since it provides quite similar results to RSM and RNG keε
models, requiring only slightly more computational power than the
standard keε model.
Fig. 11a and b shows the velocity contour and vectors through
the nozzles and crossing along the furnace calculated by the sim-
ulations. Note the swirl effect caused by primary air that shortens
the flame length.
6.2. Compressibility effects
Though the Mach number was less than 0.3, when compress-
ibility effects are considered negligible (Mach values ranging from
0.06 for natural gas burner orifice to 0.21 for secondary air burner
orifice in the present study), the present work also explored
variable-density flows in the simulation. Since upstream burner
pressure is very important in industrial burners, it was verified if
constant gas density can really be assumed. Fig. 12 shows that the
results were very similar, indicating, as expected, no practical dif-
ference between incompressible or compressible flows.
 L.C.B.S. Reis et al. / Applied Thermal Engineering 67 (2014) 201e213 207

Fig. 4. The BHF.M.10-type burner, furnace side: front view (a), side view (b).

6.3. Discharge coefficient in which Cd is the discharge coefficient, A is the geometrical orifice
area, and b is the ratio between orifice diameter and inlet diameter.
The discharge coefficient is a parameter for burner design. It is Pressure p1 is upstream from the nozzle and p2 is downstream from
defined as the ratio of the actual mass flow rate of a fluid through a the nozzle. The value of the discharge coefficient for a burner
nozzle to the ideal mass flow rate. The dependency of a volumetric nozzle is usually determined experimentally. Typically, in the
flow rate on the pressure drop of the burner can be very accurately burner industry, the discharge coefficient varies from 0.6 to 0.9.
calculated for incompressible flows by the well-known equation: This depends on several factors, including the Reynolds number of
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi the fluid in the orifice, length-to-diameter ratio of the orifice, beta
 ffi
ratio (ratio of orifice diameter to entrance pipe diameter), port
2ðp1  p2 Þ=r 1  b ;
4
Q ¼ Cd A (1)
angle (entering angle of taper) and exit angle, as shown in Fig. 13
[26]. In the case in the study, the orifices have port angle equal to

Table 1
Characteristics of the natural gas used for the simulations (typical for Usina Presidente Vargas).

CH4 C2H6 C3H8 nC4H10 C5H12 nC6H14 N2 CO2 High heat value (*) Low heat value (*) Wobbe Index (**) Gas density

Molar % kJ/m3 kJ/m3 MJ/m3 Kg/m3

89.23 5.73 1.87 0.68 0.20 0.09 0.68 1.52 39,656 35,837 45 0.74

Note: (*) m3 at 20  C and 1 atm; (**) WI calculated using the low heat value.
208 L.C.B.S. Reis et al. / Applied Thermal Engineering 67 (2014) 201e213

(a)
2500

Furnace centerline axial temperature (K)

2000

1500

fine grid coarse grid

1000

500 Furnace
Burner

0
-6 -5 -4 -3 -2 -1 0 1

Furnace centerline axial position (m)

(b)
200
NG nozzle centerline axial pressure (Pa)

150

100

50

-50
Fig. 5. Polyhedral mesh applied for the entire domain (a); a detailed view of the
polyhedral mesh applied for the burner nozzles (b). f ine grid coarse grid
-100

90 , exit angle equal to zero degrees, and ratio d/D is 0.2 for natural -150
gas nozzles. Due to these geometry conditions, it is expected that 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
those nozzles have a behavior similar to orifice plates. However, NG nozzle centerline axial position (m)
orifice plates have very short L/d ratio (Length/diameter of the
nozzle) and the nozzles being analyzed have greater L/d, equal to (c)
3.8 for natural gas nozzles.
25
Fig. 14 shows the results for calculation of the natural gas nozzle
discharge coefficients, using Equation (1) with variable values ob-
NG nozzle centerline axial velocity (m/s)

tained through the solutions from the numerical simulation. The 20

discharge coefficients are constant at approximately 0.66 for Rey-
nolds Number ranged from 25,000 to 45,000. Lower Reynolds
15
Numbers are not applicable for industrial burners. Larger L/d ratio
reduces pressure losses, increasing the discharge coefficient.
Experimental results for discharge coefficients of orifice plates 10
[1,27] have constant values of 0.61 for Reynolds Numbers above
10,000. Fig. 14 also shows a comparison with the discharge coeffi-
5
cient formula presented by Colannino [28]. Pressure losses related
to skin friction and form drag were calculated by the Idel’cik data f ine grid coarse grid
[29]: 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
!ffi
u NG nozzle centerline axial position (m)
u L X n
Cd ¼ t1= 1 þ f þ K ; (2) Fig. 6. Effect of mesh size on furnace temperature profile (a), on natural gas nozzle
D 1 pressure profile (b), and on natural gas nozzle velocity profile (c).
 L.C.B.S. Reis et al. / Applied Thermal Engineering 67 (2014) 201e213 209

Table 2a
Dimensionless distance from the wall yþ for coarse grid as a function of burner heat
loads and type of nozzle.

Nozzles Unit Burner heat load variation

Natural gas kg/s 0.057 0.068 0.080 0.085 0.093 0.101

flow rate
yþ natural gas e 108 110 116 119 125 128
Primary air kg/s 0.452 0.542 0.634 0.679 0.739 0.804
flow rate
yþ primary air e 115 141 171 181 211 238
Secondary air kg/s 0.507 0.608 0.699 0.761 0.828 0.902
flow rate
yþ secondary air e 140 148 196 214 227 245

Table 2b Fig. 7. Comparison between natural gas pressure measurements and predicted profiles
Dimensionless distance from the wall yþ for fine grid as a function of burner heat for different numerical turbulence models as a function of flow rate.
loads and type of nozzle.

Nozzles Unit Burner heat load variation

Natural gas flow rate kg/s 0.057 0.068 0.080 0.085 0.093 0.101
yþ natural gas e 57 68 80 87 99 110
Primary air flow rate kg/s 0.448 0.489 0.520 0.610 0.653 0.717
yþ primary air e 65 69 72 87 92 103
Secondary air flow rate kg/s 0.503 0.549 0.584 0.684 0.732 0.804
yþ secondary air e 56 60 63 76 80 98

In which f is the friction factor and K represents the losses of

bodies or fittings through the flow. The above relation is found
using the continuity and energy equations upstream and immedi-
ately downstream from the nozzle, including the term for frictional
losses. The discharge coefficient is constant around 0.64 regardless
the Reynolds number, agreeing well with the numerical values.
Fig. 8. Comparison between primary combustion air pressure measurements and
Another way to check the results is to use a concept presented
predicted profiles for different numerical turbulence models as a function of flow rate.
by Baukal [26] to calculate the mass flow rate through a nozzle at
subsonic flows. The equations are based on the ideal gas law as-
sumptions of ideal flow to calculate the flow rate of fuel through an

Table 3
Boundary conditions, energy and mass balance results. (a) Boundary conditions of mass flow inlet. (b) Domain outflow and mass/energy balances results.

Case# Flue gas entrance Natural gas e burner Primary air e burner Secondary air e burner

Mass flow Total heat Total sensible Mass flow Total heat Total sensible Mass flow Total heat Total sensible Mass flow Total heat Total sensible
rate transfer rate heat transfer rate transfer rate heat transfer rate transfer rate heat transfer rate transfer rate heat transfer

kg/s kW kW kg/s kW kW kg/s kW kW kg/s kW kW

1 7.02 9,080 11,141 0.101 446 0 0.804 328 328 0.902 368 368
2 7.02 9,044 11,177 0.093 409 0 0.739 301 301 0.828 338 338
3 7.02 9,020 11,202 0.085 376 0 0.679 277 277 0.761 311 311
4 7.02 8,998 11,224 0.080 351 0 0.634 259 259 0.700 285 285
5 7.02 8,861 11,361 0.068 300 0 0.542 221 221 0.608 248 248
6 7.02 8,860 11,361 0.057 250 0 0.452 185 185 0.507 207 207
Temperature: 1550 K Temperature: 300 K; nozzle diameter: Temperature: 691 K; nozzle diameter: Temperature: 691 K; nozzle diameter:
33 mm 59 mm 51 mm

Flue gas e domain exit Energy balance e domain

Case#
Mass flow Mass balance Temperature Total heat Total sensible Species mass fraction (%) Heat of Total heat Total sensible
rate net result transfer rate heat transfer reaction transfer rate heat transfer
source net result rate net result

kg/s % K kW kW O2 CO2 H2O N2 kW % %

1 8.826 0.01 1600 10,645 14,782 1.09 15.01 11.42 72.49 4,761 0.22 0.16
2 8.680 0.01 1595 10,602 14,401 1.09 15.01 11.41 72.49 4,372 0.04 0.03
3 8.542 0.03 1592 10,564 14,043 1.09 15.00 11.41 72.49 4,017 0.34 0.20
4 8.431 0.02 1590 10,550 13,772 1.07 15.02 11.43 72.48 3,754 0.45 0.30
5 8.238 0.01 1582 10,481 13,228 1.08 15.01 11.42 72.49 3,187 0.03 0.02
6 8.034 0.02 1568 10,441 12,631 1.09 15.01 11.41 72.48 2,643 0.66 0.21
Pressure outlet: 10 Pa Slab heat flux: 1,127 kW (79.45 kW/
m2); Skid heat flux: 665 kW (29.00 kW/
m2)
210 L.C.B.S. Reis et al. / Applied Thermal Engineering 67 (2014) 201e213

Fig. 9. Comparison between secondary combustion air pressure measurements and

predicted profiles for different numerical turbulence models as a function of flow rate.

orifice. To compensate for the results of the ideal equations and

assumptions, the orifice discharge coefficient is introduced to ac-
count for the condition of the real flow that makes it non-ideal. The
following equations are used to determine mass flow rate through a
nozzle:

_ ¼ Cd re AMe ce ;
m (3)

in which
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
h iffi
2=ðk  1Þ ðP =P ÞðkkÞ  1 ;
1
Me ¼ t b (4)

" #1=2
Tt kTe R P
Te ¼ ; ce ¼ ; re ¼ b :
1 þ k1 2
2 Me
MW Te R
MW

The coefficient discharge calculated by Equation (4) has the

same result as Equation (2) if Pb is considered to be equal to p2.

6.4. Temperature results

Table 4 shows a comparison between the main results of tem-

perature by means of the numerical calculation and the experi-
mental figures. Three different burner heat load conditions were
analyzed. Predicted temperatures were higher than experimental
temperatures. However, the differences are acceptable considering
the purpose of this work.
Fig. 15a shows a vertical cross-section of the temperature con-
tour along the furnace and Fig. 15b shows a horizontal cross section
of the temperature contour. Fig. 15c shows a photograph of the
flame at the same region indicated in Fig. 15a by a dashed line. This
picture was taken from the slab entrance side. It can be noted that
the shape of the simulated flame is very similar to the real one. It is
not possible to show a picture of the entire flame because the slabs
and the structural part of the skids interfere in direct view of the
flame. However, the shape and length of the flame, as shown in
Fig. 15a and b, seem to be fairly similar to the real flame for normal
operation of the burner. Fig. 15b shows that the flame has a ten-
dency to follow the streamline in the direction of the flare stack due
to flue draft, noting that there is an equal opposite burner in the
same direction.
7. Conclusion
As the design of an industrial fuel orifice is essential for a
proper mix of fuel and combustion air, an approach that is also
Fig. 10. Streamlines originating from natural gas nozzles (a), from primary combustion
air nozzles (b), and from secondary combustion air nozzles (c).
made up of the physical situation before the interaction between
fuel and oxidizer can be very important to proper combustion
modeling.
The flow through the burner nozzle has been simulated suc-
cessfully using Fluent CFD code. There is satisfactory agreement
between the numerical simulation predictions and experimental
data and also between the numerical simulation predictions and
practical approaches to burner scale. This validates the numerical
results and also the applicability of the turbulence models tested.
The keε (standard, RNG and Realizable) and RSM (Reynolds Stress
Model) models provided good results, though the standard keε
gave a slight different result from the others. The choice of the
model becomes only a matter of computational power. The Real-
izable keε is recommended considering this point.
 L.C.B.S. Reis et al. / Applied Thermal Engineering 67 (2014) 201e213 211

Fig. 11. Longitudinal cross-section of the burner/furnace: velocity contour (a), velocity vector (b).

complex situations in a very clear way and providing very valuable

recommendations.
Industrial applications with respect combustion can be sensitive
to changes in fuel gas quality. Particularly for steelworks, where
besides natural gas other fuels are used, such as coke oven gas, blast
furnace gas and steel plant gas. As those fuels are process waste
gases, variation of the chemical composition is inherent to the
process. Even for natural gas, nowadays there is a worldwide
concern regarding gas quality and consequently gas interchange-
ability, due to composition variation, in view of cross-border
transport of gas has increased a lot in recent years, and notably
the rapid developing world market in liquefied natural gas. For
industrial applications a typical variation in gas composition does
not result in operation that causes the process unsafe or inoperable.
Fig. 12. Comparison between the results for incompressible and compressible flows.
However, there are potential negative changes with respect to

It can be concluded that three dimensional numerical simula-

tions are an effective tool to design burner nozzles and can give a
more accurate burner design when used along with empirical or
practical approaches, since it is able to give results showing

Fig. 14. Natural gas nozzle discharge coefficients for differential pressure (DP1e2) from
Fig. 13. Factors influencing the discharge coefficient (length-to-diameter ratio ¼ L/d; upstream from the nozzle to downstream from the nozzle as a function of Reynolds
beta ratio ¼ d/D; port angle ¼ a, exit angle ¼ b). number.
212 L.C.B.S. Reis et al. / Applied Thermal Engineering 67 (2014) 201e213

Table 4
Comparison of experimental and simulated temperature values.

Case#2 Case#4 Case#6

Burner load: 4,372 kW Burner load: 3,754 kW Burner load: 2,643 kW

Experimental Simulation Experimental Simulation Experimental Simulation

a 
Tfurnace C 1305 1343 1290 1331 1230 1297

Tfloor C 1270 1275 1260 1274 1180 1267

Tslab-lower face C 985 1058 974 1058 926 1015

Tburner wall C 1270 1287 1260 1281 1180 1261
a
Out of the flame.

emphasize that in almost all gas appliances the flow gas is regu-
lated by making it pass through an orifice.

Acknowledgements

The authors thank Companhia Siderúrgica Nacional (CSN) for

allowing the use of its facilities and for technical support of oper-
ating personnel.

References

Fig. 15. Temperature contour along the furnace: vertical cross-section view (a), and
horizontal cross-section view (b). Photograph of the flame in the region between the
burner and the first skid (c).
emissions, efficiency and product quality. As an example, steel
reheating furnaces require controlled atmosphere to maintain the
oxygen in the flue gas under a close range, in order to ensure a
proper scale formation on the metal surface. Therefore, the use of a
numerical tool capable to redesign burner nozzles or to determine
the correct pressure for an existing burner nozzle under operation,
in case of a variation of fuel composition, is valuable to combustion
performance. For the same fuel, when there is a decrease in the
Wobbe Index, due to variation in gas composition, the only way to
adjust the delivery of the same amount of heat is increasing the
orifice upstream pressure. The Wobbe Index (WI) is a parameter of
gas interchangeability and its usefulness is that for any given
orifice, all gas mixtures that have the same WI will deliver the same
amount of heat, at the same burner pressure. It is important to
[1] M.S. Shah, J.B. Jyeshtharaj, S.K. Avtar, C.R.C. Prasad, D.S. Shukla, Analysis of
flow through an orifice meter: CFD simulation, Chem. Eng. Sci. 71 (2012) 300e
309.
[2] G.H. Nail, A Study of 3-Dimensional Flow through Orifice Meters, Ph.D.
Dissertation, Texas A&M University, USA, 1991.
[3] G.L. Morrison, R.E. DeOtte, G.H. Nail, D.L. Panak, Mean velocity and turbulence
fields inside a b ¼ 0.50 orifice flowmeter, AIChE J. 39 (1993) 745e756.
[4] C.L. Hollingshead, M.C. Johnson, S.L. Barfuss, R.E. Spall, Discharge coefficient
performance of Venturi, standard concentric orifice plate, V-cone and wedge
flow meters at low Reynolds numbers, J. Pet. Sci. Eng. 78 (2011) 559e566.
[5] N.M.B. Oliveira, L.G.M. Vieira, J.J.R. Damasceno, Numerical methodology for
orifice meter calibration, Mater. Sci. Forum 660e661 (2010) 531e536.
[6] S. Eiamsa-ard, A. Ridluan, P. Sumruasin, P. Eiasma-ard, P. Promvong, Numer-
ical investigation of turbulent flow through a circular orifice, KMITL Sci. J. 8
(2008) 43e50.
[7] A. Erdal, H.I. Andersson, Numerical aspects of flow computation through or-
ifices, Flow. Meas. Instrum. 8 (1997) 27e37.
[8] R.R. Rhinehart, S. Gebreyohannes, U.M. Sridhar, A. Patrachari, M.S. Rahaman,
A power law approach to orifice flow rate calibration, ISA Trans. 50 (2011)
329e341.
[9] W. Borutzky, B. Barnard, J. Thoma, An orifice flow model for laminar and
turbulent conditions, Simul. Model. Pract. Theory 20 (2002) 141e152.
[10] M. Reader-Harris, N. Barton, D. Hodges, The effect of contaminated orifice
plates on the discharge coefficient, Flow. Meas. Instrum. 25 (2012) 2e7.
[11] A.E. German, T. Mahmud, Modelling of non-premixed swirl burner flows
using a Reynolds-stress turbulence closure, Fuel 84 (2005) 583e594.
[12] A. Khelil, H. Naji, L. Loukarfi, G. Mompean, Prediction of a high swirled natural
gas diffusion flame using a PDF model, Fuel 88 (2009) 374e381.
[13] K.M. Saqr, H.S. Aly, M.M. Sies, M.A. Wahid, Computational and experimental
investigations of turbulent asymmetric vortex flames, Int. Commun. Heat
Mass Transfer 38 (2011) 352e362.
[14] B. Danon, E.S. Cho, W. de Jong, D.J.E.M. Roekaerts, Numerical investigation of
burner positioning effects in a multi-burner flameless combustion furnace,
Appl. Therm. Eng. 31 (2011) 3885e3896.
[15] A.O. Nieckele, M.F. Naccache, M.S.P. Gomes, Combustion performance of an
aluminum melting furnace operating with natural gas and liquid fuel, Appl.
Therm. Eng. 31 (2011) 844e851.
[16] A. Frassoldati, P. Sharma, A. Cuoci, T. Faravelli, E. Ranzi, Kinetic and fluid dy-
namics modeling of methane/hydrogen jet flames in diluted coflow, Appl.
Therm. Eng. 30 (2010) 376e383.
[17] M. Sarlej, P. Petr, J. Hájek, P. Stehlik, Computational support in experimental
burner design optimization, Appl. Therm. Eng. 27 (2007) 2727e2731.
[18] K.M. Saqr, H.S. Aly, M.M. Sies, M.A. Wahid, Effect of free stream turbulence on
NOx and soot formation in turbulent diffusion CH4 air flames, Int. Commun.
Heat Mass Transfer 37 (2010) 611e617.
[19] K. Khanafer, S.M. Aithal, Fluid-dynamic and NOx computation in swirl burners,
Int. J. Heat Mass Transfer 54 (2011) 5030e5038.
[20] H.I. Kassem, K.M. Saqr, H.S. Aly, M.M. Sies, M.A. Wahid, Implementation of the
eddy dissipation model of turbulent non-premixed combustion in Open-
FOAM, Int. Commun. Heat Mass Transfer 38 (2011) 363e367.
[21] K.M. Saqr, M.A. Wahid, Comparison of four eddy-viscosity turbulence models
in the eddy dissipation modeling of turbulent diffusion flames, Int. J. Appl.
Math. Mech. 7 (2011) 1e18.
 L.C.B.S. Reis et al. / Applied Thermal Engineering 67 (2014) 201e213
[22] T. Gassoumi, K. Guedri, R. Said, Numerical study of the swirl effect on a coaxial
jet combustion flame including radiative heat transfer, Numer. Heat Transfer
Part A 56 (2009) 897e913.
[23] M.D. Emami, A.E. Fard, Laminar flamelet modeling of a turbulent CH4/H2/N2
jet diffusion flame using artificial neural networks, Appl. Math. Model. 36
(2012) 2082e2093.
[24] Fluent Inc., Fluent 12.0.12 User’s Guide, 2009.
[25] B.E. Launder, D.B. Spalding, Mathematical Models of Turbulence, Academic
Press, London, 1972.
213
[26] C.E. Baukal, Industrial Burner Handbook, CRC Press, USA, 2003.
[27] R. Aman, H. Handroos, T. Eskola, Computationally efficient two-regime flow
orifice model for real-time simulation, Simul. Model. Pract. Theory 16 (2008)
945e961.
[28] J. Colannino, Modeling Combustion Systems: a Practical Approach, CRC Press,
Taylor & Francis Group, USA, 2006.
[29] I.E. Idel’cik, Mémento de Pertes de Charges, Eyrolles. Collection de La Direction
des Etudes etRecherchesd’Electricité de France, 1986.